blob: 3f72b87ee74755c09bc6a47629a977be2ea46c81 [file] [log] [blame]
# RUN: llc -mtriple=aarch64-linux-gnu -mcpu=falkor -run-pass aarch64-falkor-hwpf-fix-late -o - %s | FileCheck %s
---
# Verify that the tag collision between the loads is resolved for various load opcodes.
# CHECK-LABEL: name: hwpf1
# CHECK: $[[BASE:[a-z0-9]+]] = ORRXrs $xzr, $x1, 0
# CHECK: LDRWui $[[BASE]], 0
# CHECK: LDRWui $x1, 1
name: hwpf1
tracksRegLiveness: true
body: |
bb.0:
liveins: $w0, $x1
$w2 = LDRWui $x1, 0 :: ("aarch64-strided-access" load (s32))
$w2 = LDRWui $x1, 1
$w0 = SUBWri $w0, 1, 0
$wzr = SUBSWri $w0, 0, 0, implicit-def $nzcv
Bcc 9, %bb.0, implicit $nzcv
bb.1:
RET_ReallyLR
...
---
# CHECK-LABEL: name: hwpf2
# CHECK: $[[BASE:[a-z0-9]+]] = ORRXrs $xzr, $x1, 0
# CHECK: LD1i64 $q2, 0, $[[BASE]]
# CHECK: LDRWui $x1, 0
name: hwpf2
tracksRegLiveness: true
body: |
bb.0:
liveins: $w0, $x1, $q2
$q2 = LD1i64 $q2, 0, $x1 :: ("aarch64-strided-access" load (s32))
$w2 = LDRWui $x1, 0
$w0 = SUBWri $w0, 1, 0
$wzr = SUBSWri $w0, 0, 0, implicit-def $nzcv
Bcc 9, %bb.0, implicit $nzcv
bb.1:
RET_ReallyLR
...
---
# CHECK-LABEL: name: hwpf3
# CHECK: $[[BASE:[a-z0-9]+]] = ORRXrs $xzr, $x1, 0
# CHECK: LD1i8 $q2, 0, $[[BASE]]
# CHECK: LDRWui $x1, 0
name: hwpf3
tracksRegLiveness: true
body: |
bb.0:
liveins: $w0, $x1, $q2
$q2 = LD1i8 $q2, 0, $x1 :: ("aarch64-strided-access" load (s32))
$w0 = LDRWui $x1, 0
$w0 = SUBWri $w0, 1, 0
$wzr = SUBSWri $w0, 0, 0, implicit-def $nzcv
Bcc 9, %bb.0, implicit $nzcv
bb.1:
RET_ReallyLR
...
---
# CHECK-LABEL: name: hwpf4
# CHECK: $[[BASE:[a-z0-9]+]] = ORRXrs $xzr, $x1, 0
# CHECK: LD1Onev1d $[[BASE]]
# CHECK: LDRWui $x1, 0
name: hwpf4
tracksRegLiveness: true
body: |
bb.0:
liveins: $w0, $x1
$d2 = LD1Onev1d $x1 :: ("aarch64-strided-access" load (s32))
$w2 = LDRWui $x1, 0
$w0 = SUBWri $w0, 1, 0
$wzr = SUBSWri $w0, 0, 0, implicit-def $nzcv
Bcc 9, %bb.0, implicit $nzcv
bb.1:
RET_ReallyLR
...
---
# CHECK-LABEL: name: hwpf5
# CHECK: $[[BASE:[a-z0-9]+]] = ORRXrs $xzr, $x1, 0
# CHECK: LD1Twov1d $[[BASE]]
# CHECK: LDRWui $x1, 0
name: hwpf5
tracksRegLiveness: true
body: |
bb.0:
liveins: $w0, $x1
$d2_d3 = LD1Twov1d $x1 :: ("aarch64-strided-access" load (s32))
$w0 = LDRWui $x1, 0
$w0 = SUBWri $w0, 1, 0
$wzr = SUBSWri $w0, 0, 0, implicit-def $nzcv
Bcc 9, %bb.0, implicit $nzcv
bb.1:
RET_ReallyLR
...
---
# CHECK-LABEL: name: hwpf6
# CHECK: $[[BASE:[a-z0-9]+]] = ORRXrs $xzr, $x1, 0
# CHECK: LDPQi $[[BASE]]
# CHECK: LDRWui $x1, 3
name: hwpf6
tracksRegLiveness: true
body: |
bb.0:
liveins: $w0, $x1
$q2, $q3 = LDPQi $x1, 3 :: ("aarch64-strided-access" load (s32))
$w0 = LDRWui $x1, 3
$w0 = SUBWri $w0, 1, 0
$wzr = SUBSWri $w0, 0, 0, implicit-def $nzcv
Bcc 9, %bb.0, implicit $nzcv
bb.1:
RET_ReallyLR
...
---
# CHECK-LABEL: name: hwpf7
# CHECK: $[[BASE:[a-z0-9]+]] = ORRXrs $xzr, $x1, 0
# CHECK: LDPXi $[[BASE]]
# CHECK: LDRWui $x1, 2
name: hwpf7
tracksRegLiveness: true
body: |
bb.0:
liveins: $w0, $x1
$x2, $x3 = LDPXi $x1, 3 :: ("aarch64-strided-access" load (s32))
$w2 = LDRWui $x1, 2
$w0 = SUBWri $w0, 1, 0
$wzr = SUBSWri $w0, 0, 0, implicit-def $nzcv
Bcc 9, %bb.0, implicit $nzcv
bb.1:
RET_ReallyLR
...
---
# Verify that the tag collision between the loads is resolved and written back
# for post increment addressing for various load opcodes.
# CHECK-LABEL: name: hwpfinc1
# CHECK: $[[BASE:[a-z0-9]+]] = ORRXrs $xzr, $x1, 0
# CHECK: LDRWpost $[[BASE]], 0
# CHECK: $x1 = ORRXrs $xzr, $[[BASE]], 0
# CHECK: LDRWui $x1, 1
name: hwpfinc1
tracksRegLiveness: true
body: |
bb.0:
liveins: $w0, $x1
$x1, $w2 = LDRWpost $x1, 0 :: ("aarch64-strided-access" load (s32))
$w2 = LDRWui $x1, 1
$w0 = SUBWri $w0, 1, 0
$wzr = SUBSWri $w0, 0, 0, implicit-def $nzcv
Bcc 9, %bb.0, implicit $nzcv
bb.1:
RET_ReallyLR
...
---
# CHECK-LABEL: name: hwpfinc2
# CHECK: $[[BASE:[a-z0-9]+]] = ORRXrs $xzr, $x1, 0
# CHECK: LD1i64_POST $q2, 0, $[[BASE]]
# CHECK: $x1 = ORRXrs $xzr, $[[BASE]], 0
# CHECK: LDRWui $x1, 1
name: hwpfinc2
tracksRegLiveness: true
body: |
bb.0:
liveins: $w0, $x1, $q2
$x1, $q2 = LD1i64_POST $q2, 0, $x1, $x1 :: ("aarch64-strided-access" load (s32))
$w2 = LDRWui $x1, 132
$w0 = SUBWri $w0, 1, 0
$wzr = SUBSWri $w0, 0, 0, implicit-def $nzcv
Bcc 9, %bb.0, implicit $nzcv
bb.1:
RET_ReallyLR
...
---
# CHECK-LABEL: name: hwpfinc3
# CHECK: $[[BASE:[a-z0-9]+]] = ORRXrs $xzr, $x1, 0
# CHECK: LD1i8_POST $q2, 0, $[[BASE]]
# CHECK: $x1 = ORRXrs $xzr, $[[BASE]], 0
# CHECK: LDRWui $x1, 132
name: hwpfinc3
tracksRegLiveness: true
body: |
bb.0:
liveins: $w0, $x1, $q2
$x1, $q2 = LD1i8_POST $q2, 0, $x1, $x1 :: ("aarch64-strided-access" load (s32))
$w0 = LDRWui $x1, 132
$w0 = SUBWri $w0, 1, 0
$wzr = SUBSWri $w0, 0, 0, implicit-def $nzcv
Bcc 9, %bb.0, implicit $nzcv
bb.1:
RET_ReallyLR
...
---
# CHECK-LABEL: name: hwpfinc4
# CHECK: $[[BASE:[a-z0-9]+]] = ORRXrs $xzr, $x1, 0
# CHECK: LD1Rv1d_POST $[[BASE]]
# CHECK: $x1 = ORRXrs $xzr, $[[BASE]], 0
# CHECK: LDRWui $x1, 252
name: hwpfinc4
tracksRegLiveness: true
body: |
bb.0:
liveins: $w0, $x1, $q2
$x1, $d2 = LD1Rv1d_POST $x1, $xzr :: ("aarch64-strided-access" load (s32))
$w2 = LDRWui $x1, 252
$w0 = SUBWri $w0, 1, 0
$wzr = SUBSWri $w0, 0, 0, implicit-def $nzcv
Bcc 9, %bb.0, implicit $nzcv
bb.1:
RET_ReallyLR
...
---
# CHECK-LABEL: name: hwpfinc5
# CHECK: $[[BASE:[a-z0-9]+]] = ORRXrs $xzr, $x1, 0
# CHECK: LD3Threev2s_POST $[[BASE]]
# CHECK: $x1 = ORRXrs $xzr, $[[BASE]], 0
# CHECK: LDRWroX $x17, $x0
name: hwpfinc5
tracksRegLiveness: true
body: |
bb.0:
liveins: $w0, $x1, $x17, $q2
$x1, $d2_d3_d4 = LD3Threev2s_POST $x1, $x0 :: ("aarch64-strided-access" load (s32))
$w0 = LDRWroX $x17, $x0, 0, 0
$w0 = SUBWri $w0, 1, 0
$wzr = SUBSWri $w0, 0, 0, implicit-def $nzcv
Bcc 9, %bb.0, implicit $nzcv
bb.1:
RET_ReallyLR
...
---
# CHECK-LABEL: name: hwpfinc6
# CHECK: $[[BASE:[a-z0-9]+]] = ORRXrs $xzr, $x1, 0
# CHECK: LDPDpost $[[BASE]]
# CHECK: $x1 = ORRXrs $xzr, $[[BASE]], 0
# CHECK: LDRWui $x17, 2
name: hwpfinc6
tracksRegLiveness: true
body: |
bb.0:
liveins: $w0, $x1, $x17, $q2
$x1, $d2, $d3 = LDPDpost $x1, 3 :: ("aarch64-strided-access" load (s32))
$w16 = LDRWui $x17, 2
$w0 = SUBWri $w0, 1, 0
$wzr = SUBSWri $w0, 0, 0, implicit-def $nzcv
Bcc 9, %bb.0, implicit $nzcv
bb.1:
RET_ReallyLR
...
---
# CHECK-LABEL: name: hwpfinc7
# CHECK: $[[BASE:[a-z0-9]+]] = ORRXrs $xzr, $x1, 0
# CHECK: LDPXpost $[[BASE]]
# CHECK: $x1 = ORRXrs $xzr, $[[BASE]], 0
# CHECK: LDRWui $x17, 2
name: hwpfinc7
tracksRegLiveness: true
body: |
bb.0:
liveins: $w0, $x1, $x17, $q2
$x1, $x2, $x3 = LDPXpost $x1, 3 :: ("aarch64-strided-access" load (s32))
$w18 = LDRWui $x17, 2
$w0 = SUBWri $w0, 1, 0
$wzr = SUBSWri $w0, 0, 0, implicit-def $nzcv
Bcc 9, %bb.0, implicit $nzcv
bb.1:
RET_ReallyLR
...
---
# Check that we handle case of strided load with no HW prefetcher tag correctly.
# CHECK-LABEL: name: hwpf_notagbug
# CHECK-NOT: ORRXrs $xzr
# CHECK: LDARW $x1
# CHECK-NOT: ORRXrs $xzr
# CHECK: LDRWui $x1
name: hwpf_notagbug
tracksRegLiveness: true
body: |
bb.0:
liveins: $w0, $x1, $x17
$w1 = LDARW $x1 :: ("aarch64-strided-access" load (s32))
$w1 = LDRWui $x1, 0 :: ("aarch64-strided-access" load (s32))
$w17 = LDRWui $x17, 0 :: ("aarch64-strided-access" load (s32))
$w0 = SUBWri $w0, 1, 0
$wzr = SUBSWri $w0, 0, 0, implicit-def $nzcv
Bcc 9, %bb.0, implicit $nzcv
bb.1:
RET_ReallyLR
...
---
# Check that we treat sp based loads as non-prefetching.
# CHECK-LABEL: name: hwpf_spbase
# CHECK-NOT: ORRXrs $xzr
# CHECK: LDRWui $x15
# CHECK: LDRWui $sp
name: hwpf_spbase
tracksRegLiveness: true
body: |
bb.0:
liveins: $w0, $x15
$w1 = LDRWui $x15, 0 :: ("aarch64-strided-access" load (s32))
$w17 = LDRWui $sp, 0
$w0 = SUBWri $w0, 1, 0
$wzr = SUBSWri $w0, 0, 0, implicit-def $nzcv
Bcc 9, %bb.0, implicit $nzcv
bb.1:
RET_ReallyLR
...
---
# Check that non-base registers are considered live when finding a
# scratch register by making sure we don't use $x2 for the scratch
# register for the inserted ORRXrs.
# CHECK-LABEL: name: hwpf_offreg
# CHECK: $x3 = ORRXrs $xzr, $x1, 0
# CHECK: $w10 = LDRWroX $x3, $x2, 0, 0
name: hwpf_offreg
tracksRegLiveness: true
body: |
bb.0:
liveins: $w0, $x1, $x2, $x17, $x18
$w10 = LDRWroX $x1, $x2, 0, 0 :: ("aarch64-strided-access" load (s32))
$x2 = ORRXrs $xzr, $x10, 0
$w26 = LDRWroX $x1, $x2, 0, 0
$w0 = SUBWri $w0, 1, 0
$wzr = SUBSWri $w0, 0, 0, implicit-def $nzcv
Bcc 9, %bb.0, implicit $nzcv
bb.1:
RET_ReallyLR
...